Uniswap Driver
Warn
Audited by Gen Agent Trust Hub on Nov 1, 2026
Risk Level: MEDIUMMETADATA_POISONINGEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [METADATA_POISONING]: The skill's manifest and description claim it is "Provided by Uniswap" and use Uniswap's official homepage and logo. However, the actual author is identified as "bankrbot". This discrepancy can mislead users into believing the skill is an official product supported by the service provider.
- [EXTERNAL_DOWNLOADS]: The installation process involves executing "npx skills add Uniswap/uniswap-ai", which downloads and configures code from an external repository. While the target repository is associated with a well-known service, this mechanism introduces a dependency on external code execution at installation time.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to research market conditions and verify tokens on-chain to inform its link generation, creating an attack surface for instructions embedded in external data.
- Ingestion points: Market condition research and on-chain token verification (SKILL.md).
- Boundary markers: No specific delimiters or safety instructions are provided in the manifest to separate untrusted data from the agent's core logic.
- Capability inventory: Generates deep links for the Uniswap web interface (catalog.json).
- Sanitization: There is no evidence of content filtering or validation for the data retrieved from external sources before it is used to generate user-facing links.
Audit Metadata